Show HN: Rampart – Runtime firewall for Claude Code and AI agents in YOLO mode (github.com)

Rampart has been launched as a runtime firewall specifically designed for Claude Code and other AI agents operating in YOLO mode. This tool offers enhanced visibility and control over AI operations by evaluating every tool call against pre-defined security policies before execution. With Rampart, potentially dangerous commands are blocked within microseconds, and all decisions are logged in a hash-chained audit trail, ensuring integrity and traceability of actions taken by AI agents. The tool is significant for the AI/ML community as it addresses a critical safety concern—allowing AI agents to manage infrastructure unsupervised can lead to risks associated with erroneous or malicious command execution. Rampart provides seamless integration with existing systems through native hooks or syscall interception and supports customizable YAML-based policies to define what actions are allowed or denied. Its capability to assess most decisions rapidly, with additional LLM-based classification for ambiguous commands, enhances security without imposing significant performance overhead. The comprehensive logging and notification features further assist organizations in maintaining security compliance and oversight in their AI deployments.